Introduction
In this course, I propose a computational analysis of my personal playlist on Spotify. My pocket playlist comprises disco, dance-pop, and sentimental ballads, which I have been accumulating since September 2020. These songs are looped or appeared in my pocket list at least for 2 weeks, which proves that they are my favourites during my university life.
The collection of these songs is solely influenced by my YouTube explore, Spotify recommendations and radio shows. Since my background allows me to access different language-pop over the Asia, the playlist is composed round 90 songs from languages, such as Korean, English, Mandarin, Cantonese, Taiwanese, and Japanese. The diversity of my favourites is inspirational to my project and I hope to understand the mechanism behind my choice.
The analysis would be conducted according to the geographical region/verbal language of the playlist: United States (English), Korea (Korean), Hong Kong (Cantonese), Taiwan (Mandarin/Taiwanese) and Japan (Japanese). Each region has its distinctive features, such as: danceability, acousticness, valence and tempo; however, each connects each with the combination of its genre and languages, and this project aims to evaluate their similarities and patterns. Thus, in each category, both typical and unique tracks are selected, to represent or emphasize their typicality/atypicality.
Notes:
The written words of Cantonese, Mandarin and Taiwanese are all based on Chinese, in different accents.
Language as Genre: Danceability
Table 1: Danceability distribution in different languages
English has the highest medium of danceability among all genres, followed by Korean. Both Cantonese and Taiwanese have the even distribution of danceability between 0.4 and 0.8. Mandarin and Japanese have the lower medium of danceability among all genres. Overall, all tracks are higher than 0.5, which symbolizes my tendency in listening music with high dancebility
Language as Genre: Tempo
Table 2: Tempo distribution in different languages
Vice versa, the distribution in tempo shows an opposite pattern comparing to danceability, where English has the lowest medium, excluding atypical tracks: Starboy, Blinding Lights (The Weeknd) and Stay (The Kid Laroi ft. Justin Bieber). Along with Cantonese, both distribution are the narrowest between 100 and 130 BPM. Taiwanese has the highest medium of tempo at 150 BPM. Overall, most of the medium of tempo distribution locate around 125 BPM.
Language as Genre: Valence
Table 3: Valence Distribution in different languages
The majority of distribution have higher tendency of valence, referring that my playlist possesses relatively music positivity, such as Cantonese, Japanese and Tawiwanese. However, the tendency of Mandarin music goes opposite, whereas there are higher density at 0.3. It illustrates the music negativity of Mandarin tracks in my playlist, which correlates to the sentimental ballads I have chosen.
Histogram of Key distribution
Table 4: Key distribution in different languages
The key of C and C#/Db have the highest counts of tracks in my playlist, which English composes most of them. The key of D#/Eb and E appear as the least count where Mandarin composes most of them. It also shows the key distribution between languages, such as Korean has the evenest distribution.
Comparison: Chroma features of The Weeknd
Table 5: Chromagram of “Blinding Light” (The Weeknd)
From table 5, we can see that F key has a consistent pattern of its timbre feature in “Blinding Light”. We might hardly recognize the chorus on the chromagram; however tracing back to the track, a strong rhythmic echo in F key clearly stands out as the bass of the song. We can also see a cycle throughout the entire song, where F key goes to C, then C# and keep circulating.
Table 6: Chromagram of “In Your Eyes” (The Weeknd)
From table 6, the euclidean timbre features show the route of the bass, from G to C to F and back to G. It is relatively obvious compared to above, with many sudden change between chords, and it is not the chorus but the bass of the track.
By comparing these 2 songs from the same artist, we can see that C and F chords often appear as the key timbre.
Self-Similarity Matrix of “I Can’t Stop me”
Table 7: Timbre of “I Can’t Stop Me” (Twice)
From table 7, it displays a few blue chequerboard and yellow outlines. Such as between 40-60 secs, 100-120 secs and 160-180 secs, the chequerboards are identicial, symbolizing the pre-chorus of track. The yellow grid lines appear when human voices exist; vice versa, such as 60-80 and 170-190 secs, these are the chorus with pure instrumental.
Table 8: Pitches of “I Can’t Stop Me” (Twice)
From table 8, we can see that the pitch class of “I Can’t Stop Me” are consistent throughout the track, with densed diagonal lines. The yellow grid lines clearly divide the track into parts. If we zoom into 0-5 secs, 35-40 secs, 75 secs, 100 secs, 140-150 secs and 190-200 secs, these are the moments/periods where pitches transpose.
Overall, we can analyze the structure of “I Can’t Stop Me”: Intro -> Verse -> Pre-chorus -> chorus -> Verse -> Pre-chorus -> chorus -> Bridge -> pre-chorus -> chorus -> outro [ABABCB].
((i cant show both diagram on the same page as it overlaps, problem to be solved))
Chordogram of ‘Yoru ni kakeru (Run to the night)’
Table 9: Chordogram of “Yoru ni kakeru 夜に駆ける” (Run to the night)
We can see two obvious yellow vertical lines at 15-30 and 105-120 secs, these are the chorus without human voices. If we zoom into the bridge between 195-205 secs,we can see that few of the blue boxes transpose to Bmin/Dmaj/F#min/Amaj while two of yellow boxes transpose to Ebmaj/Abmaj. According to the track, the bridge and ending chorus have transposed to a higher key. It might nor clearly state in the chordogram, but we can see the blue box transpose from C minor to Cmaj/Amin/Fmaj/Dmin (actually the track reaches A minor).
Standard deviation
